“Backtesting,” “paper trading,” and “demo account” get used almost interchangeably by beginners, but they test genuinely different things, at different points in developing a strategy. Confusing them is a common reason traders think they’ve “tested” a strategy when they’ve really only tested one part of it.
The Short Version
- Backtesting runs your strategy’s rules against historical price data to see how it would have performed in the past. Fast, statistical, but based on data that already happened.
- Paper trading and demo accounts are effectively the same idea under different names — placing simulated orders with fake money against current, live (or near-live) market conditions. “Paper trading” is more common in stock and crypto platform terminology; “demo account” is the standard term in forex and CFD trading. Slow, real-time, but based on data that’s happening now.
Neither replaces the other — they answer different questions.
Backtesting: What It Actually Tests
Backtesting takes a strategy you can express as a set of rules — “buy when X happens, sell when Y happens” — and applies those rules mechanically to historical price data, producing a hypothetical track record: win rate, average gain/loss, drawdowns, and so on.
What it’s good for:
- Quickly testing whether a rules-based idea has any historical edge at all, across years of data in minutes.
- Comparing variations of a strategy (different stop-loss levels, different timeframes) far faster than you could by trading each variation live.
- Spotting how a strategy performs across different market regimes (trending vs. choppy, high vs. low volatility) if you have enough historical data to cover both.
What it consistently gets wrong:
- Overfitting: it’s easy to tweak a strategy’s rules until they fit the historical data extremely well — and just as easy for that fit to be coincidence rather than a real edge. A strategy overfit to the past routinely fails going forward.
- No execution reality: a backtest generally assumes your orders fill at exactly the historical price, instantly, with no slippage — an even more optimistic assumption than most paper trading tools make.
- Survivorship and data quality issues: backtests run on incomplete or adjusted historical data (for example, missing delisted stocks) can quietly overstate how well a strategy would have actually performed.
- It can’t test discretionary decisions: if your strategy involves judgment calls that can’t be written as an explicit rule, there’s nothing for a backtest to mechanically apply.
Paper Trading / Demo Accounts: What They Actually Test
Paper trading and demo accounts place simulated trades against real-time market data, as it happens — you’re trading the actual market conditions of today, just with fake money instead of real. See our full explanation of how the mechanics work for more detail on order simulation specifically.
What it’s good for:
- Testing a strategy prospectively, without the risk that you unconsciously shaped its rules to fit data you’d already seen (a subtler version of the overfitting problem above).
- Practicing the actual mechanics of placing orders, managing open positions, and reacting to live price movement in real time — none of which a backtest exercises at all.
- Getting a feel (even if imperfect) for real-time decision-making under uncertainty, since you don’t know what happens next the way you do when replaying history.
What it consistently gets wrong:
- Psychology, still: fake money doesn’t create the same stress as real money, so discipline that holds up in paper trading can still collapse once real capital is at risk.
- Slow feedback: testing a strategy meant for daily timeframes might realistically take months of paper trading to gather a statistically meaningful sample — far slower than a backtest covering years of data in minutes.
- Optimistic fills: as covered in our simulator mechanics guide, many paper trading tools still assume better execution than you’d get live, just like backtests do, only with real-time data instead of historical.
Using Them Together: A Reasonable Workflow
- Backtest first to quickly filter out strategies with no historical edge at all, and to compare structural variations (stop-loss placement, timeframe, entry filters) efficiently.
- Paper trade the survivors to confirm the strategy still behaves sensibly in current, live conditions, and to build the actual execution habits (order entry, position management) a backtest can’t teach you.
- Watch for a gap between the two. If a strategy performed well historically but struggles in paper trading under current conditions, that’s a meaningful signal the market regime may have shifted, or that the backtest was more overfit than it looked.
- Only then consider real money, starting small — and even then, treat the transition itself as something to prepare for deliberately; see our forex demo-to-live readiness checklist for a framework that applies beyond forex specifically.
Frequently Asked Questions
Is backtesting more reliable than paper trading?
They’re reliable at different things. Backtesting is more statistically efficient (years of data in minutes) but more prone to overfitting and unrealistic execution assumptions. Paper trading is slower but tests real, current conditions and actual order-placement mechanics that a backtest skips entirely.
Can I skip backtesting and just paper trade?
You can, especially for discretionary strategies that can’t be cleanly expressed as backtestable rules. You’ll lose the ability to quickly test historical performance across many market conditions, so expect it to take longer to build confidence in the strategy either way.
Why did my backtested strategy fail in paper trading?
The most common reasons are overfitting to historical data, unrealistic fill assumptions in the backtest, or a genuine shift in market conditions between the historical period tested and now. Treat a live gap like this as useful information about the strategy’s real robustness, not just bad luck.
Do demo accounts and paper trading use different data than backtesting?
Yes — demo accounts and paper trading use current, real-time (or near real-time) market data, while backtesting uses historical data, sometimes going back years. This is the core reason they answer different questions rather than being two versions of the same test.